By using this site, you agree to our Privacy Policy and our Terms of Use. Close

Forums - Nintendo Discussion - Wii U GPU Type CONFIRMED! Custom AMD E6760!

HoloDust on 24 September 2012
Baron said:
HoloDust said:
 

5000 score is for GPU score, P score is 5600+ (against 5870 in spec sheet of e6760), so I'd say that's pretty close.
As for 4850 - if you look you can find it by yourself, but here are 2 results at stock speeds:

http://www.3dmark.com/3dmv/3144570

http://www.3dmark.com/3dmv/3574315

That's some 1.3x of stated P score of e6760

Now, I much more prefer using pure GPU scores, as they are not that dependant of CPU - take a look at this result for 4850, but with Phenom II 955 and see that GPU score is pretty much the same as with 620

http://www.3dmark.com/3dmv/3843371

That said, if you want to persist in claiming that e6760 is not based on Turks with 480:24:8 configuration, that's your choice, if not feel free to look for GPU scores of 6570 and 6650m at stock speeds (which vary from 4500-5300) and compare them to 4850 GPU score of around 7500.

So you believe there can be up to 3000 points difference between the same setups without any clock improvements? Come on...

As I said before, 3dmark doesn't always record the correct speeds.

Here's the 4850 with a 50% faster cpu, the QX9650;
http://nl.hardware.info/productinfo/benchmarks/6/3d-chips?products[]=26815&specId=3912&tcId=132

Here's the 4850 with a Core i7 965 which is about 130% faster than the Athlon II X4 620;
http://nl.hardware.info/productinfo/benchmarks/6/3d-chips?tcId=190&specId=4935&products[]=26815

These are independent benchmarks with every component at stock speeds.

Now are you really gonna try and say that a 4850 with a 620 can do the same scores as a 4850 with a Core i7 965?

From what I see the 4850 scores 6100 points at best. Just a few hundred points above the e6760, or a few percent.

In any way, there's no way a 4850 scores well over 6000 points without an overclock or a much faster processor. Let alone 7500+.

 

So no, the 4850 isn't 1.6 or 1.4 or 1.3 times faster. Not even close.

3DMark results (stock clocks):
QX9650 + 4850 http://www.3dmark.com/3dmv/3337424

i7965 + 4850 http://www.3dmark.com/3dmv/3874142

Now, notice difference in P score (not much), notice difference in CPU scores (huge) and notice difference in GPU scores (almost non-existant). Compare GPU scores (7335 and 7441) with GPU results for Athlon II 620 and Phenom 955 (7332 /7395 and 7513). Now, do you get what I'm talking about? You notice how P score when you go from lowest CPU (620) to highest (i7-965) is slighlty raising (7608-8727)? You notice how CPU score is wildly raising (8578-20256)? Now, notice how GPU score is almost identical in all cases? So that's 4850 3DMark GPU Vantage score (not P). Now compare that with 6570/6650M/6750M GPU scores at stock speeds (somewhere around 5200-5300). Divide 7400 (on average) with 5300...around 1.4...Or divide any of stock speed 620/4850 combo results (which are around 7600) with e6760 score from official specs...or just use logic and common sense and compare 4850 specs with e6760 specs....whaterever you do I got bit tired of explaining that card with 800:40:16 config with 10GPixel/s-25GTexel/s-63.55GB/s memory bandiwth with 1000GFLOPs performance is quite a bit more powerfull than card with 480:24:8 config with 4.8GPIxel/s-14.4GTexel/s-51.2GB/s bandwith with 576GFLOPS. So excuse me if I don't continue to elaborate on this matter anymore in the future...



Around the Network
Baron on 24 September 2012
HoloDust said:

3DMark results (stock clocks):
QX9650 + 4850 http://www.3dmark.com/3dmv/3337424

i7965 + 4850 http://www.3dmark.com/3dmv/3874142

Now, notice difference in P score (not much), notice difference in CPU scores (huge) and notice difference in GPU scores (almost non-existant). Compare GPU scores (7335 and 7441) with GPU results for Athlon II 620 and Phenom 955 (7332 /7395 and 7513). Now, do you get what I'm talking about? You notice how P score when you go from lowest CPU (620) to highest (i7-965) is slighlty raising (7608-8727)? You notice how CPU score is wildly raising (8578-20256)? Now, notice how GPU score is almost identical in all cases? So that's 4850 3DMark GPU Vantage score (not P). Now compare that with 6570/6650M/6750M GPU scores at stock speeds (somewhere around 5200-5300). Divide 7400 (on average) with 5300...around 1.4...Or divide any of stock speed 620/4850 combo results (which are around 7600) with e6760 score from official specs...

Why? An e6760 is not a 6570 or a 6650m or a 6750m. I thought that was quite clear already. What's the point of releasing a more expensive embedded solution when customers can just go and buy a 6750m when that's the same thing?

And your still using 3dmark results which have absolutely no validity as we can't say whether or not the speeds recorded are the speeds used to run the benchmark. I've showed you two results that are undeniably real no matter how many 3dmark user results you throw at it. Your claim that there's no big difference in total score between different levels of cpu power is absolutely bullshit. Anyone who has ever used different processors in Vantage can tell you that. More cpu power is higher score, A lot more cpu power is a much higher score. That is just fact.

or just use logic and common sense and compare 4850 specs with e6760 specs....whaterever you do I got bit tired of explaining that card with 800:40:16 config with 10GPixel/s-25GTexel/s-63.55GB/s memory bandiwth with 1000GFLOPs performance is quite a bit more powerfull than card with 480:24:8 config with 4.8GPIxel/s-14.4GTexel/s-51.2GB/s bandwith with 576GFLOPS.

So it's common sense to say that, based on specs, one card is a lot more powerful but it's not common sense to look at the results those cards produce and conclude they perform the same? All right then.

On paper it's quite a bit more powerful but as the results I provided show, when it comes to performance they are quite comparable. I wonder how that could be.... Perhaps being embedded directly into the motherboard has its advantages?...

Food for thought there.



HoloDust on 24 September 2012
Baron said:
HoloDust said:
 

3DMark results (stock clocks):
QX9650 + 4850 http://www.3dmark.com/3dmv/3337424

i7965 + 4850 http://www.3dmark.com/3dmv/3874142

Now, notice difference in P score (not much), notice difference in CPU scores (huge) and notice difference in GPU scores (almost non-existant). Compare GPU scores (7335 and 7441) with GPU results for Athlon II 620 and Phenom 955 (7332 /7395 and 7513). Now, do you get what I'm talking about? You notice how P score when you go from lowest CPU (620) to highest (i7-965) is slighlty raising (7608-8727)? You notice how CPU score is wildly raising (8578-20256)? Now, notice how GPU score is almost identical in all cases? So that's 4850 3DMark GPU Vantage score (not P). Now compare that with 6570/6650M/6750M GPU scores at stock speeds (somewhere around 5200-5300). Divide 7400 (on average) with 5300...around 1.4...Or divide any of stock speed 620/4850 combo results (which are around 7600) with e6760 score from official specs...

Why? An e6760 is not a 6570 or a 6650m or a 6750m. I thought that was quite clear already. What's the point of releasing a more expensive embedded solution when customers can just go and buy a 6750m when that's the same thing?

And your still using 3dmark results which have absolutely no validity as we can't say whether or not the speeds recorded are the speeds used to run the benchmark. I've showed you two results that are undeniably real no matter how many 3dmark user results you throw at it. Your claim that there's no big difference in total score between different levels of cpu power is absolutely bullshit. Anyone who has ever used different processors in Vantage can tell you that. More cpu power is higher score, A lot more cpu power is a much higher score. That is just fact.

or just use logic and common sense and compare 4850 specs with e6760 specs....whaterever you do I got bit tired of explaining that card with 800:40:16 config with 10GPixel/s-25GTexel/s-63.55GB/s memory bandiwth with 1000GFLOPs performance is quite a bit more powerfull than card with 480:24:8 config with 4.8GPIxel/s-14.4GTexel/s-51.2GB/s bandwith with 576GFLOPS.

So it's common sense to say that, based on specs, one card is a lot more powerful but it's not common sense to look at the results those cards produce and conclude they perform the same? All right then.

On paper it's quite a bit more powerful but as the results I provided show, when it comes to performance they are quite comparable. I wonder how that could be.... Perhaps being embedded directly into the motherboard has its advantages?...

Food for thought there.

Hm, wonder have you looked at test date of those "undeniably real" results? October 2010...Care to find something fresher with newer drivers? And for user results...I gave you several that are around same area at stock speeds....so  I suppose you're saying they are invalid cause they don't fit your case...Right....

Now, if you cared to actually read what I wrote, you could've seen that P score does rise with different CPUs, albeit way more midly than CPU scores. On the other hand, GPU score is around the same for 4850 at stock speeds (7300-7500) no matter what CPU. But those are invalid too, cause....wait, they don't fit your case...Right...

And as for embeded....this is not Apple, there's nothing "magical" about it - in case of e6760 embedded means that GPU and memory are in one package, on same die...with 51.2GB/s bandwith...oh wait, that's the same bandwith as with 6750m with 800Mhz clock for memory (same as e6760)...no, no,  it's not the same chip as 6570/6650m/6750m, it's completely something different...Right...



drkohler on 24 September 2012
HoloDust said:

And as for embeded.... in case of e6760 embedded means that GPU and memory are in one package, on same die...

Nonsense. If you bought the embedded version, you would get GPU and memory (and whatever else is required) on a single BGA. Of course memory and GPU are on completely separate dies. I highly doubt the WiiU is a standard solution (since the GPU is heavily customized to start with), though.



HoloDust on 24 September 2012
drkohler said:
HoloDust said:

And as for embeded.... in case of e6760 embedded means that GPU and memory are in one package, on same die...

Nonsense. If you bought the embedded version, you would get GPU and memory (and whatever else is required) on a single BGA. Of course memory and GPU are on completely separate dies. I highly doubt the WiiU is a standard solution (since the GPU is heavily customized to start with), though.


Yeah, my bad, that's what I meant under embedded for default e6760.



Around the Network
FrancisNobleman on 24 September 2012

if we went back to the old nomenclature of bits, what would the wii u be ?

I miss the old school 8, 32, 128 bits. it was so cool and simple.



Viper1 on 24 September 2012
FrancisNobleman said:
if we went back to the old nomenclature of bits, what would the wii u be ?

I miss the old school 8, 32, 128 bits. it was so cool and simple.


Hard to really say because there are various parts of the system with different bit depths.   Consoles since the PS2, GC and Xbox have been 32/64/128/256 bit mixed systems.  



The rEVOLution is not being televised

zarx on 25 September 2012

Ok official word from AMD

“Kevin we can confirm that the Wii U will use AMD graphics but we have not provided any specifics on what silicon they will be using in the upcoming console.”

http://gimmegimmegames.com/2012/09/amd-rep-says-wii-u-uses-custom-radeon-e6760-gpu/



@TheVoxelman on twitter

Check out my hype threads: Cyberpunk, and The Witcher 3!

Deoz on 25 September 2012
zarx said:

Ok official word from AMD

“Kevin we can confirm that the Wii U will use AMD graphics but we have not provided any specifics on what silicon they will be using in the upcoming console.”

http://gimmegimmegames.com/2012/09/amd-rep-says-wii-u-uses-custom-radeon-e6760-gpu/


And there it is ladies and gentleman.



HoloDust on 25 September 2012
zarx said:

Ok official word from AMD

“Kevin we can confirm that the Wii U will use AMD graphics but we have not provided any specifics on what silicon they will be using in the upcoming console.”

http://gimmegimmegames.com/2012/09/amd-rep-says-wii-u-uses-custom-radeon-e6760-gpu/


This is getting weirder by the hour - another update on that story (Update 2 in link you posted), the original tech support guys confirms e6760:

http://www.neogaf.com/forum/showpost.php?p=42495681&postcount=5388

Since this rumour has been out for quite a while (I've seen some back in July), I'm starting to think there's really something to it - customized off course, and it does not nesecerally mean it's at stock speed, I've even seen someone claiming it's clocked at around 800Mz which would put it into 6670 teritory. Anyway, not that long and we'll know for sure, I'm much more interested to see what that rumoured CPU is all about, and will it turn out to be bottleneck for some games.